On Minimal Covolume Hyperbolic Lattices

We study lattices with a non-compact fundamental domain of small volume in hyperbolic space Hn. First, we identify the arithmetic lattices in Isom+Hn of minimal covolume for even n up to 18. Then, we discuss the related problem in higher odd dimensions and provide solutions for n = 11 and n = 13 in terms of the rotation subgroup of certain Coxeter pyramid groups found by Tumarkin. The results depend on the work of Belolipetsky and Emery, as well as on the Euler characteristic computation for hyperbolic Coxeter polyhedra with few facets by means of the program CoxIter developed by Guglielmetti. This work complements the survey about hyperbolic orbifolds of minimal volume.


Introduction
Let n ≥ 2 and consider a hyperbolic lattice, that is, a discrete group Γ ⊂ Isom H n whose orbit space or orbifold Q = H n /Γ is of finite volume.By a celebrated result of Kazhdan and Margulis, the set of all volumes vol n (Q) has a positive minimal element µ n .In the work [1], we provided a survey about the values µ n and the volume minimising n-orbifolds for dimensions n satisfying 2 ≤ n ≤ 9 by taking into account (non-)compactness, orientability, arithmeticity, and dimension parity.
In this work, we consider only the volumes of non-compact or cusped hyperbolic n-orbifolds and study the corresponding volume spectrum with minimal element ν n .The set V n contains the proper subset V a n of volumes of orientable quotient spaces of H n by arithmetic lattices in Isom + H n with corresponding minimal element ν a n .By deep results of Belolipetsky and Emery (see [2][3][4][5]), the values ν a n are explicitly known for n ≥ 4. Our aim is to describe the hyperbolic lattices whose covolumes equal ν a n for n ≥ 10.In this context, hyperbolic lattices generated by finitely many reflections in hyperplanes of H n , called hyperbolic Coxeter groups, are of particular interest (see Section 2.3).In fact, for n ≤ 9, the smallest covolume hyperbolic Coxeter n-simplex groups (generated by n + 1 reflections) are all arithmetic and yield the unique non-compact volume minimisers (up to a factor of two, in the exceptional case n = 7; for references, see [1], Section 3).In this way, the values ν n and ν a n could be entirely specified.However, in Isom H n , cofinite Coxeter simplex groups do not exist for n ≥ 10 and, apart from Borcherds' example [6] for n = 21, nothing is known about the existence of cofinite hyperbolic Coxeter groups for n ≥ 20.
In the sequel of our commensurability classification of Coxeter pyramid groups with n + 2 generators existing up to dimension 17 (see [7]), Guglielmetti [8] developed the software program CoxIter testing various properties such as arithmeticity and providing invariants such as the Euler characteristic of a hyperbolic Coxeter group.In Section 2.6, we give several instructive examples.By a result of Emery [9], the covolume of the single Coxeter pyramid group Γ * ⊂ Isom H 17 with Coxeter graph given by Figure 1 yields the minimal value among all ν a n for n ≥ 2 (see also Section 3.1.2).Based on these facts, we are able to identify the orientable cusped arithmetic hyperbolic n-orbifolds as orbit spaces by the action of certain hyperbolic Coxeter groups for the even dimensions n with 10 ≤ n ≤ 18 and for the odd dimensions n = 11 and n = 13 (the proof for n = 13 is based on a combinatorial argument due to S. Tschantz [10]).The results are presented in Proposition 1, Proposition 2 and Proposition 3 of Section 3.1.The work ends with a couple of remarks about the mysterious case of dimension n = 15 and the non-arithmetic case.

Hyperbolic n-Space
Denote by X n one of the standard geometric spaces given by either the Euclidean space E n , the standard sphere S n , or hyperbolic space H n .View each space X n in the context of a linear space equipped with a suitable bilinear form •, • .In particular, H n is interpreted as a connected component H n of the two-sheeted hyperboloid in R n+1 according to: In this picture, the isometry group Isom H n is isomorphic to the group PO(n, 1) of positive Lorentzian matrices leaving invariant the product •, • n,1 and H n (cf.[11] Chapter 3).By passing to the upper half space model U n of H n in E n + , the line element and the volume element are given by: , where a > 0, carries a Euclidean metric up to a distortion factor 1/a 2 .Such a subset is called a horosphere based at ∞ and bounds the horoball B ∞ (a) = { x ∈ U n | x n > a } above it.

Cusped Hyperbolic Orbifolds of Finite Volume
Consider a discrete subgroup Γ ⊂ Isom H n with fundamental domain D ⊂ H n , and suppose that Γ is of finite covolume or cofinite for short.This means that the volume of the orbifold Q = H n /Γ, as given by the volume of D, is finite, and we call Γ a hyperbolic lattice, for short.In the sequel we are particularly interested in lattices Γ giving rise to non-compact or cusped orbifolds H n /Γ.Such a group Γ contains at least one non-trivial subgroup Γ q of parabolic type whose elements stabilise a point q ∈ ∂H n , say (see [12], Section 3.1).Associated to the fixed point q at infinity is a cusp neighborhood C q ⊂ V which is an embedded subset given by the quotient B q /Γ q of a Γ q -precisely invariant horoball B q in H n .The action of Γ q on the boundary horosphere ∂B q is by Euclidean isometries and with a compact fundamental domain so that Γ q can be interpreted as a crystallographic subgroup of Isom E n−1 .By Bieberbach's theory, any crystallographic group in Isom E n−1 contains a finite index translational lattice of rank n − 1.For n ≤ 9, it is known (see [13,14] and also Section 3) that small volume cusped hyperbolic n-orbifolds are intimately related to dense lattice packings in E n−1 .

Hyperbolic Coxeter Polyhedra and Discrete Reflection Groups with Few Generators
Consider a geometric polyhedron P ⊂ X n , that is, P is an n-dimensional convex polyhedron of finite volume in X n , bounded by N ≥ n + 1 hyperplanes H i ⊂ X n with unit normal vector e i directed away from, say, P. Denote by G(P) = ( e i , e j ) 1≤i,j≤N , the Gram matrix of P. In [15,16], Vinberg developed a very satisfactory theory to conclude the existence and describe arithmetic, combinatorial, and metrical properties of P in terms of G(P).There are explicit criteria by means of certain submatrices of G(P) allowing one to decide whether P is compact or of finite volume.In this work, we are dealing mainly with non-compact hyperbolic polyhedra of finite volume, being the convex hull of finitely many ordinary points in H n and at least one point in the boundary ∂H n at infinity of H n .In fact, the fundamental group of a finite volume cusped hyperbolic n-orbifold admits a fundamental domain whose closure is a hyperbolic polyhedron and any of its parabolic subgroups stabilises a vertex at infinity of the fundamental polyhedron.
A geometric polyhedron P ⊂ X n is a Coxeter polyhedron if all of its dihedral angles α = α F formed by pairs H, H of intersecting hyperplanes in the boundary of P, and with associated ridge F = H ∩ H ∩ P, are of the form π/m for an integer m ≥ 2. Coxeter polyhedra arise in a natural way as building blocks in the context of regular polyhedra and as closures of fundamental regions of discrete reflection groups.Denote by Γ = Γ(P) the group generated by the N reflections s = s H with respect to the hyperplanes H bounding the Coxeter polyhedron P; the group Γ is called the geometric Coxeter group associated to P. The group Γ is a discrete subgroup in Isom X n which admits a particularly simple presentation with relations satisfying s 2 = 1 and for distinct generators s = s H and s = s H , with hyperplanes H and H intersecting in X n along a ridge H ∩ H ∩ P where P has dihedral angle π/m.Most conveniently, geometric Coxeter polyhedra of simple combinatorics (and Coxeter groups with few generators) are described by their Coxeter graph Σ.Each node ν in Σ corresponds to a hyperplane H (and the reflection s = s H ) and is joined to another node ν by an edge with weight m if the corresponding dihedral angle, formed by their hyperplanes H, H at the ridge F in P, equals α F = π/m.Usually, edges with a weight 2 are omitted and edges with weight 3 (resp.4) are drawn as simple (resp.double edges).In the hyperbolic case, and for parallel hyperplanes H, H intersecting in ∂H n , the nodes ν, ν are connected by a bold edge; for hyperplanes H, H disjoint in H n and of hyperbolic distance l, the nodes ν, ν are connected by a dotted edge (and sometimes marked with the weight l).
In contrast to the spherical and Euclidean cases, Coxeter polyhedra in H n are classified only very partially.There is a complete list for hyperbolic Coxeter simplices, characterised by N = n + 1, and they exist for n ≤ 9, only.Hyperbolic Coxeter polyhedra with N = n + 2 are classified and they exist for n ≤ 17.Notice that examples of compact Coxeter polyhedra in H n are known just for n ≤ 8.In higher dimensions, there are only single examples in H n for n = 18, 19, and 21 whose discovery is due to Kaplinskaja, Vinberg, and Borcherds, respectively.Notice that Coxeter polyhedra in H n do not exist for n > 995.For a survey, we refer to information and references collected on the webpage of Felikson and Tumarkin [17].
As for even dimensions above 17, there are only the two Coxeter polyhedra P 18 and P L 18 explicitly known (for more details, see Example 7).The polyhedron P 18 exists in H 18 and is non-compact and bounded by 37 hyperplanes only forming dihedral angles of π/2 and π/3.The polyhedron was discovered by Kaplinskaja and Vinberg [18] and is associated with the maximal reflection group Γ 18 in the group PO(18, 1; Z) of integral positive linear transformations leaving invariant the unimodular quadratic form q n = x, x n,1 for n = 18.Observe that the quadratic forms q n are reflective in the above sense, providing finite volume non-compact hyperbolic Coxeter polyhedra P n , for n ≤ 19.More precisely, the group PO(n, 1; Z) is the automorphism group PO(I n,1 ) of the odd unimodular Lorentzian lattice I n,1 with quadratic form q n whose maximal reflection subgroup is of finite index equal to the order of the symmetry group Sym(P n ) of its (fundamental) Coxeter polyhedron P n .The order of Sym(P n ) is different from 1 for 14 ≤ n ≤ 19 and equal to 2 (resp.4) for n = 14, 15 (resp.n = 16, 17), whereas the symmetric group S n appears according Sym(P 18 ) ∼ = S 4 and Sym(P 19 ) ∼ = S 5 .The results can be found in [18,19] and ( [16] part II, Chapter 6, §2).

Arithmetic Hyperbolic Coxeter Groups
The group PO(18, 1; Z) is a model of an arithmetic group, a notion which will not be explained in detail here (see [4,5], for example).One characterisation is -by using a result of Margulis (see [20], Theorem 10.3.5, for example) -that a lattice G ⊂ Isom H n , n ≥ 3 , is non-arithmetic if and only if its commensurator group Comm(G) is discrete in Isom H n (and containing G with finite index).Here, the group Comm(G) is defined by: However, for hyperbolic Coxeter groups Γ ⊂ PO(n, 1) such as Γ 18 , Vinberg developed a very useful criterion for arithmeticity.This criterion simplifies drastically when the group Γ has a non-compact (fundamental) Coxeter polyhedron P ⊂ H n .Consider the Gram matrix G = G(P) of P and form the matrix Theorem (Vinberg's Criterion).Let P ⊂ H n be a non-compact hyperbolic Coxeter polyhedron with Coxeter group Γ and Gram matrix G.Then, Γ is arithmetic if and only if all the cycles of the matrix 2 • G are rational integers.
Example 1.The non-compact hyperbolic Coxeter simplices with graphs Ξ n , 2 ≤ n ≤ 9 , given in Table 1 are all arithmetic.

The Euler Characteristic and the Covolume of a Hyperbolic Coxeter Group
Let Γ ⊂ Isom H n be a Coxeter group with presentation S | R according to (1) and fundamental polyhedron P ⊂ H n of finite volume.Consider the growth series where l S (γ) is the word length of γ ∈ Γ with respect to the generating set S of Γ. Denote F = { T ⊂ S | Γ T < Γ finite } the set of all subsets T of S such that the group Γ T generated by the elements in T is finite.Notice that the groups of type Γ T are spherical Coxeter groups with finite growth series.In order to represent their growth polynomials, we use the standard notations and so on, and denote by m 1 = 1, m 2 , . . ., m t the exponents of the Coxeter group Γ T (see [22], Section 9.7).For the list of irreducible finite Coxeter groups, see Table 2.

Graph Exponents
Growth polynomial f T (x) By a result of Steinberg [23], f S (x) is the power series of a rational function and satisfies the following important formula. 1 where Γ ∅ = {1} .For the Euler characteristic χ(Γ), one obtains, for any abstract Coxeter group Γ, In terms of the volume of P and therefore the quotient space H n /Γ , one deduces the following identity (see [24]).
The formulas (3) and ( 5) are very useful when computing the volume of an even-dimensional hyperbolic Coxeter polyhedron.Since the list of irreducible finite Coxeter groups is known and comparatively short (see Table 2), this volume computation can be realised by a computer program.

The Computer Program CoxIter
By means of the computer program CoxIter designed by Guglielmetti [8] in 2015 (freely accessible online https://coxiter.rgug.ch/,https://coxiterweb.rafaelguglielmetti.ch/), different invariants of a Coxeter group Γ acting by reflections on H n can be computed.The input are the dimension n and the Coxeter graph Σ with the number of its nodes and with the edge weights m > 2 (either 0 or −1 for parallel or disjoint hyperplanes, respectively).Then, the program CoxIter answers questions such as cocompactness, cofiniteness, arithmeticity, Euler characteristic and covolume of Γ, number of vertices at infinity, and the f -vector (with components f i equal to the number of i-dimensional faces) of its Coxeter polyhedron P. Example 3. Consider the two Coxeter pyramids P 10 ⊂ H 10 and P 12 ⊂ H 12 with Coxeter graphs given by Figures 2 and 3.By the tools mentioned in Sections 2.4 and 2.5, one can check easily that the associated Coxeter groups Γ 10 and Γ 12 are arithmetic.By means of the webversion of CoxIter, one computes their Euler characteristic as being equal to χ(Γ 10 ) = −1/183936614400 (see also [14], Appendix A3) and χ(Γ 12 ) = 691/382588157952000 (see the output given in Figure 4).Example 4. Consider the Coxeter group Γ 14 generated by 17 reflections in Isom H 14 with graph given in Figure 5.It was discovered by Vinberg as being the maximal reflection subgroup of the group of units of the unimodular quadratic form q 14 of signature (14,1).The program CoxIter yields the following information.6, also discovered by Vinberg, is the maximal reflection subgroup of the group of units of the unimodular quadratic form q 16 .Here, CoxIter provides the following data (see also [8], Table 2).
In an earlier work, together with Kaplinskaja, Vinberg [18] used the algorithm mentioned above to prove that the unimodular quadratic forms q n are also reflective for n = 18 and n = 19 (while this is not the case for n ≥ 20).Furthermore, they provided the corresponding Coxeter graphs.By CoxIter, Guglielmetti computed the covolume of the Coxeter group Γ 18 ⊂ Isom H 18 , related to q 18 , which is generated by 37 reflections, and found that χ(Γ 18 ) = − 109638854849 1482580623111880900608000000 so that the covolume is ≈ 2.204424 × 10 −12 (see [8] Table 4).
Observe that the numerator 109638854849 = 691 × 3617 × 43867 of χ(Γ 18 ) is identical to the one of χ(Γ L 18 ), a direct consequence of the formula (4) and the fact that both Coxeter graphs have integer weights 2 and 3, only.

Minimal Volume Cusped Hyperbolic Orbifolds
Let Γ ⊂ Isom H n be a lattice with fundamental polyhedron P ⊂ H n such that the n-orbifold H n /Γ is non-compact.This implies that P is the convex hull of finitely many vertices, with at least one vertex q belonging to ∂H n , whose stabiliser Γ q ⊂ Γ is a crystallographic group (see Section 2.2).Consider the volume spectrum: of all cusped hyperbolic n-orbifolds together with its minimal element ν n ∈ V n for each n ≥ 2.
In [12], a universal lower volume bound for cusped hyperbolic n-manifolds has been established that also holds in the singular case (see [28] Section 2, [13] Section 2 and [14] Chapter 5).It provides a lower bound for ν n in terms of (lattice) packing densities and orders of maximal point groups.
More precisely, denote by vol • k the Euclidean k-volume functional and by B(r) a Euclidean r-ball.Let ϕ k be the maximal point group order of elements in a fixed Q-class of maximal, finite, absolutely irreducible subgroups of GL(k, Z), and let δ k be the maximal lattice packing density in Euclidean k-space.In particular, one has that ϕ 24 = 2 22 × 3 9 × 5 4 × 7 2 × 11 × 13 × 23 = 8315553613086720000, which is equal to twice the order of the Conway group Co 1 , and that δ 8 = π 4 /384 ≈ 0.25367 and δ 24 = π 12 /12!≈ 0.00193 (see [29]).Notice that for n ≤ 8, the densest lattice packings are known and intimately related to root lattices.Moreover, by a recent fundamental result of Viazovska [30], the E 8 root lattice yields the densest sphere packing in E 8 , leading to a proof of similar flavour, by Cohn, Kumar, Miller, Radchenko and Viazovska [31], showing that the Leech lattice is an optimal sphere packing in E 24 .
The value d n (∞) denotes the simplicial horoball density, that is, the density of n + 1 horoballs based at the vertices of an ideal regular simplex in H n .By means of the volume ω n of an ideal regular n-simplex with its representation as an infinite series given by Milnor, d n (∞) can be expressed as follows (see [12], Theorem 2.1). where The first ten values of d n (∞) and d 25 (∞) are given in Table 3. Theorem (Kellerhals [12,28]).Let n ≥ 2, and let Q = H n /Γ be a hyperbolic n-orbifold with m ≥ 1 cusps.Then, As an example, by using the data of the Leech lattice in (8), the volume of a 25-dimensional hyperbolic orbifold with m ≥ 1 cusps can be bounded from below as follows.
The questions about the explicit value and the realisation of the minimal volume ν n as vol n (H n /Γ • ) have only partial answers so far.By the above theorem, one deduces the bound ν n ≥ c n for n ≥ 2, which is a key ingredient in answering the question for n ≤ 9.In fact, the classical results for the dimensions n = 2, due to Siegel, and n = 3, due to Meyerhoff, were extended by Hild-Kellerhals [28] for n = 4 and by Hild [13,14] for n ≤ 9, with the consequence that, for these dimensions, the unique covolume minimising groups Γ • ⊂ Isom H n are given by certain hyperbolic Coxeter groups (up to index two in dimension n = 7).For a survey, see [1].It turns out that all these groups Γ • ⊂ Isom H n , n ≤ 9 , are arithmetic and related to a tessellation of H n by a 1-cusped Coxeter simplex.

The Arithmetic Case
In view of the situation just described and when looking to dimensions n ≥ 10, it makes sense to study the (proper) subset V a n ⊂ V n of all volumes of orientable cusped hyperbolic n-orbifolds with arithmetic fundamental groups and to ask corresponding questions about the minimal element in V a n , denoted by ν a n > 0.
For an arbitrary dimension n, there is the standard arithmetic group PO(I n,1 ) of automorphisms that leave the form q n invariant.This group provides a first candidate for small volume.As already mentioned, for n ≤ 19, the group PO(I n,1 ) is reflective and can be written as the semi-direct product of its cofinite maximal reflection subgroup and the symmetry group Sym(P n ) of its (fundamental) Coxeter polyhedron P n .For the covolume of PO(I n,1 ), one has the following result for n even.

|χ(PO(I
with a plus sign if n ≡ 0, 2 mod 8 and the minus sign if n ≡ 4, 6 mod 8.

Even Dimensions
In a much more general context, Belolipetsky and Emery (see [2][3][4][5] and [9]) successfully exploited a relevant structural result of Prasad and determined the explicit value of ν a n for the cases of orientable cusped arithmetic orbifolds of even dimensions n ≥ 4 and of odd dimensions n ≥ 5, respectively (notice that non-compactness is not a constraint in their works).In particular, for even dimensions, there is the following result in terms of the Euler characteristic.
While the evaluation of Belolipetsky's theorem for even dimensions 4 ≤ n ≤ 8 coincides with the previously mentioned (and more explicit) results of Kellerhals and Hild (see [28] and [14]), it yields the following Table 4 for even dimensions 10 ≤ n ≤ 18.
Let us compare the results in Belolipetsky's theorem with the values χ(Γ n ) , 10 ≤ n = 2r ≤ 18 , and χ(Γ L 18 ) obtained in Section 2.6.Some of the Coxeter group examples presented in Section 2.3 have Coxeter graphs admitting a non-trivial symmetry group S k of order k, say, which corresponds to the symmetry group of the same order of the associated Coxeter polyhedron.By extending the Coxeter group by S k , we pass to a group of 1/k-times the covolume of the original group.Furthermore, since reflections are orientation reversing isometries, we need to pass to the index two orientation preserving subgroup.By taking into account the uniqueness property in Belolipetsky's result, we can deduce the following explicit volume minimality result.

Proposition 1.
Let n be even with n ∈ {10, 12, 14, 16, 18}.Then, the unique orientable cusped arithmetic hyperbolic n-orbifold Q n = H n /∆ n is given by the action on H n of the index two orientation preserving subgroup ∆ n of the group Θ n ⊂ Isom H n given by Table 5.For odd dimensions n ≤ 9, the results of Hild provide a complete picture about minimal volume cusped hyperbolic n-orbifolds, arithmetically defined or not, including proofs for uniqueness, a presentation of the fundamental group, and the value of ν n .The orbifolds are closely related to Coxeter simplices, which do not exist for dimensions n ≥ 10 (see Example 1 and Table 1).Combinatorially very close are pyramids over a product of two simplices of positive dimensions, which have been studied and classified in the Coxeter case by Tumarkin (see Section 2.3).These groups are generated by n + 2 reflections, they are all non-compact and exist in Isom H n for all 4 ≤ n ≤ 17 with n = 14, 15, 16.By Vinberg's arithmeticity criterion (see Section 2.4), one verifies easily their arithmeticity when n ≥ 11.According to the corresponding commensurability classification performed in [7], one has five Coxeter pyramid groups in Isom H 11 falling into two commensurability classes, three Coxeter pyramid groups in Isom H 13 forming one commensurability class, and finally the single Coxeter pyramid group Γ * ⊂ Isom H 17 that is closely related to the automorphism group of the even unimodular group PO(II 17,1 ) (see Example 2).Among the five arithmetic Coxeter pyramid groups Isom H 11 , which fall into two commensurability classes, the group Γ 11 given by the graph in Figure 9 has smallest covolume, and among the three commensurable Coxeter pyramid groups in Isom H 13 , the group Γ 13 given by Figure 11 has smallest covolume (see [7] and [32]).In order to identify explicitly-if possible-the minimal volume orientable cusped arithmetic hyperbolic n-orbifolds for n ≥ 11 odd, we provide details of the corresponding result of Belolipetsky and Emery (see Section 3.1.1).
(1) If r ≡ 1 (mod 4): In [9], Emery described in more detail the fundamental group ∆ n of the orientable arithmetic cusped orbifold Q n of minimal volume as follows.For n ≡ 1 (mod 8), the group PSO(II n,1 ) is conjugate to ∆ n in Isom H n , while for n ≡ 5 (mod 8) the group PSO(I n,1 ) is conjugate to a subgroup of index 3 of ∆ n in Isom H n .For n ≡ 3 (mod 4), the group ∆ n is commensurable to the group PO( f 3 ; Z) of integral automorphisms of the Lorentzian form of signature (n, 1) given by: By a result of Mcleod [33], the group PO( f 3 ; Z) is reflective for n ≤ 13.As an extension of their work for PO(n, 1; Z) to the group PO( f d ; Z), Ratcliffe and Tschantz determined the covolumes of the groups PO( f 3 ; Z) and, for each n ≡ 3 (mod 4), they computed furthermore the commensurability ratio κ n ∈ Q of ∆ n and PO( f 3 ; Z) showing that κ n = 1 (see [34], (35)).
In view of these results and the knowledge of hyperbolic Coxeter group candidates in Isom H n for n = 11, 13 and n = 17, we provide the following new characterisation of Q n = H n /∆ n for n = 11, 13 and mention briefly the known result of Emery [9] in the case n = 17.
The case n = 11.Since 11 ≡ 3 (mod 4), the group ∆ 11 of minimal covolume is commensurable to the group PO( f 3 ; Z), whose cofinite maximal reflection subgroup Γ was described by Mcleod In particular, by the result ([34], Table 1), one gets the value covol 11 (PO( where L(s, D) is Dirichlet's L-function according to ( [34], ( 12)), as well as κ 11 = 1 4 (2 5 − 1) (2 6 + 1) = 2015 4 (see [34], Section 7).This implies that vol 11 (P) =  [10] when comparing the corresponding Coxeter graphs, there is a close combinatorial relation between the polyhedron P associated to Mcleod's Coxeter group Γ and the Coxeter pyramid P 11 .In fact, pass to the double P σ of the polyhedron P by reflecting it in the bounding hyperplane depicted by σ in the graph of Figure 10.Then, the polyhedron P σ is bounded by 16 hyperplanes.Reflect recursively the pyramid P 11 in its facets while staying inside the polyhedron P σ .The image pyramids match along their facets or line up with the facets of P σ .It takes exactly 4030 copies of the pyramid P 11 to fill P σ .As a consequence and by (12) (13).
The case n = 13.Since 13 satisfies n ≡ 5 (mod 8), the fundamental group ∆ n of the orientable arithmetic cusped orbifold Q n of minimal volume is commensurable to the special unimodular group PSO(I 13 ), with ratio of their covolumes equal to 3 (see [9], Proposition 5).By the result [27], Theorem 6, of Ratcliffe and Tschantz mentioned in Remark 1, the covolume of PSO(I 13 ), being of index two in PO(I 13 ), can be expressed as follows.Proposition 3. The orientable arithmetic cusped hyperbolic 13-orbifold Q 13 of minimal volume is the quotient of H 13 by the rotation subgroup Γ + 13 of Γ 13 .Its volume ν a 13 is given by (14).
The case n = 17.Let us finish by mentioning the result [9], Theorem 2, of Emery.It states that for n ≡ 5 (mod 8), the minimal volume orientable arithmetic cusped hyperbolic n-orbifold Q n is the quotient space H n /PSO(II n,1 ).For n = 17, the group PO(II 17,1 ) is reflective and is the semi-direct product of the reflection group Γ * with the symmetry group S 2 of P * , where P * is Tumarkin's Coxeter pyramid with graph given in Figure 1 and described in Example 2. By exploiting the theorem above, one gets the following volume identification (see [9]

Final remarks.
(1) When looking at realisations of orbifolds with volumes equal to the minimal values ν a n for n ≤ 18, there remains a need to study the case n = 15 and to look for a candidate in the commensurability class of PO( f 3 ; Z) that is conjugate to the fundamental group of the minimal volume hyperbolic orbifold of dimension 15.
(2) It is an interesting but difficult question whether, and to what extent, non-arithmetic considerations can perturb the picture described in Section 3 in such a way that ν a n > ν n for some n > 3.

dim n Ξ n 2 b b b 3 b b b b 6 4 b b b b b 5 b b b b b b 6 b b b b b b b 7 b b b b b b b b 8 b b b b b b b b b 9 b b b b b b b b b b Example 2 .
The Coxeter polyhedron P * ⊂ H17 given by the graph in Figure1is bounded by 19 hyperplanes and has precisely two vertices at infinity.It has the combinatorial type of a pyramid over a product of two (eight-dimensional) simplices.Coxeter polyhedra of this type have been classified by Tumarkin[21].The polyhedron P * yields an arithmetic reflection group Γ * , as is easily checked by means of Vinberg's criterion above.The group Γ * is the maximal reflection in the automorphism group PO(II 17,1 ) of the even unimodular Lorentzian lattice II 17,1 .Due to the obvious two-fold symmetry of the graph, one can pass to the Z 2 -extension of the group Γ * , which is arithmetic of half the covolume of Γ * .

Figure 4 .
Figure 4. Output of the webversion of CoxIter for the group Γ 12 .

Table 2 .
Exponents and growth polynomials of irreducible finite Coxeter groups Γ T .

Table 5 .
The groups Θ n .